Microsoft brings five AI video models to Foundry through fal partnership
The public preview pairs text- and image-to-video options with per-second pricing. Microsoft says the lineup includes the first US-hosted Kling endpoint.
Loading page…
The public preview pairs text- and image-to-video options with per-second pricing. Microsoft says the lineup includes the first US-hosted Kling endpoint.
Listen to this story
The Nov. 5 agreement gives Microsoft Foundry customers a curated set of five fal-served video models as a first-party Azure service, rather than access to fal’s full catalog of more than 1,500 models. The options cover text-to-video, image animation and audio, with listed rates varying by model, resolution and sound settings. The launch is limited to public preview: US hosting does not guarantee every organization’s data requirements are met, and customers must check each model’s terms and supported production capabilities.
The lineup includes Kling v3 Pro, LTX-2.5 Fast and Pro, and MiniMax H3 Max and H3 Max Turbo; capabilities include synchronized audio and reference-to-video for H3 Max.
At 1080p, listed rates are $0.08 per second for MiniMax H3 Max Turbo, $0.13 for LTX-2.5 Fast and $0.17 for LTX-2.5 Pro; Kling ranges from $0.1232 to $0.2156 depending on audio settings.
Microsoft says the initial models are US-hosted, but directs customers to review each model’s processing and storage terms and verify its supported governance and security capabilities.
Developers building video into business applications have five more model options coming to Microsoft Foundry. Microsoft and fal announced the partnership on November 5, with the models rolling out in public preview as a first-party Azure service served by fal. The lineup covers generating video from written instructions and animating existing images.
Microsoft’s starting point is that video workloads need different tools. A team exploring a written creative brief may want several quick interpretations. A retailer animating an approved product image has a different starting point. The partnership is intended to let developers choose models for those steps and connect them with other capabilities in Foundry.
fal provides APIs—interfaces developers use to connect software—to more than 1,500 models across images, video, audio, 3D and generative worlds. This agreement brings a selected set of video models into Foundry, rather than that entire catalog. fal serves the models, which are sourced from the companies that build them.
Microsoft lists MiniMax H3 Max Turbo at $0.08 per generated video second at 1080p.
Microsoft lists LTX-2.5 Fast at $0.13 per generated video second at 1080p.
Microsoft lists LTX-2.5 Pro at $0.17 per generated video second at 1080p.
Microsoft’s launch pricing table labels its rates in dollars per generated video second. At 480p, MiniMax H3 Max starts at $0.05 and H3 Max Turbo at $0.025. At 720p, LTX-2.5 Fast is $0.09 and Pro is $0.12. Those starting prices cover different resolutions, so they are not a like-for-like comparison.
Higher-resolution output costs more within the listed options. MiniMax H3 Max reaches $0.16 per second at 1080p, while Turbo reaches $0.08. LTX-2.5 Fast extends to 1440p at $0.19 and 4K at $0.30; the table lists Pro at 720p and 1080p.
For Kling v3 Pro, the listed price changes with audio settings: $0.1232 per video second with audio off, $0.1848 with audio on, and $0.2156 with audio plus voice control. Sound is therefore a separate budget choice within that offering, not simply a feature included at one rate.
Microsoft says the initial portfolio is hosted in the United States and that Foundry is the first provider to offer a US-hosted Kling endpoint. That is a hosting claim, not a blanket assurance that every organization’s data requirements are met. Microsoft tells customers to review each model’s documented processing and storage terms.
For teams moving toward production, Microsoft points to evaluation, optimization, governance, security, monitoring and infrastructure capabilities supported by their selected offering. The qualification matters: customers need to check the support attached to the model they choose, rather than treating the five-model portfolio as one uniform service specification.
The rollout remains a public preview. Microsoft says more models are planned over time and that pricing will be added to its Azure pricing page in the coming weeks. Those are the next stated steps; the announcement does not give a date for general availability.
Loading discussion...
Join the conversation
Explain when generated sound would be worth the extra cost.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.